Hacker News Reader

Comment by xnickb | Hacker News Reader

I would recommend not using pixelation as means to hide text. It is easy to reverse.

Personally, if I wanted something to not be read by anyone I would never leave it in the source image and pixel blur it out. That doesn't make sense. Why leave the original text there at all? Cut the entire area out so there is no text, and nothing that can be reversed in that cast. I'd even take it a step further and say that once you cut the text out, screenshot the image and use that instead of the edited original, that way even embedded edit/undo/smart os features can't accidentally leak out the ability to undo to the original.

Funny. I was doing the same guided by some gut feeling, and then the Pixel vulnerability was published, where people would recover original images from the images cropped in Pixel album app

Is it? I hear that thrown around a lot but I doubt it's easy. Maybe people are confusing it will a swirly filter, which isn't destructive and can be reversed easily?

I know that there's strong prior knowledge but the pixelation is destructive so the problem is very ill-posed.

Fair enough, I haven't tried it myself, just glanced over a piece of software that does it (HMM) and saw the blogpost.

In reality , though, removing just a bit pf entropy can be of great value when context is known. With LLMs in mind.

It's not necessarily easy, but possible in some cases. If you can can live with a risk of, say, 1% that someone reverses this, fine. If not, I recommend using black bars.

> pixelation is destructive

Yes, but JPEG compression is destructive too yet you can (well, most of the time) see what it's in the picture.

With pixelation you are essentially replace a clearly readable character with some yet unknown 'pixel character'. It's even more pronounced on a fixed-width fonts.

Just try it yourself on the clear shot (not a JPEGed to death one) first.

Reply on news.ycombinator.com